Abstract:Single particle tracking has long been utilized for investigation of complex system dynamics such as nanoparticle-cell interaction, and the analysis of complicated behaviors of individual particles is always an outstanding issue. Existing methods treat each data point or short fragment on the recorded trajectory as an isolated atom and determine their relationship by using externally predefined mechanical models or physical states, which inevitably lead to oversimplification of spatiotemporal complexity. Herein, inspired by the phenomenology study in philosophy, we propose a machine-learning-based modeless framework to analyze particle trajectories. This new strategy consists of 3 steps: 1) make the atomic data-points on the entire trajectory all connected by constructing successive overlapped historical vectors; 2) perform unsupervised clustering in the huge vector space to find their relative differences; 3) project the differences back to the trajectory for data visualization. Consequently, the inner heterogeneity of the spatiotemporal data object emerges on its own as a colored trajectory, exhibiting both a global picture and the details on local transitions of the particle movement, without the need to predefine any models, parameters or training sets. We demonstrated with simulated data that the underlying mechanism of various time-series of particle motion could be well resolved in a consistent manner. Experimentally, we collected time-series data of lateral positions and orientations of gold nanorods, and demonstrated that our analysis method could reveal the motion states and detailed dynamics of the nanoparticles during their entire transmembrane process. Since it is becoming increasingly difficult to find mechanical-model-based solutions for complexity studies at various scales, and the dynamic behaviors of individual molecules, biological cells, cars and even stars could all be reduced to spatiotemporal trajectories of single particles, this presuppositionless approach could help the fundamental researches on many important systems.

Connecting the dots across time: Reconstruction of single cell signaling trajectories using time-stamped data

Inference of Spatiotemporal Effects on Cellular State Transitions from Time-Lapse Microscopy

Beyond pseudotime: Following T-cell maturation in single-cell RNAseq time series

Recovering biomolecular network dynamics from single-cell omics data requires three time points

Inferring Population Dynamics from Single-Cell RNA-sequencing Time Series Data.

Concepts and Limitations for Learning Developmental Trajectories from Single Cell Genomics.

Single-cell morphodynamical trajectories enable prediction of gene expression accompanying cell state change

Trajectory inference from single-cell genomics data with a process time model

Reconstructing gene regulatory dynamics from high-dimensional single-cell snapshot data

Reconstructing Disease Dynamics for Mechanistic Insights and Clinical Benefit

Reconstructing signaling history of single cells with imaging-based molecular recording

Multi-Experiment Nonlinear Mixed Effect Modeling of Single-Cell Translation Kinetics after Transfection

Dynamic Microfluidic Cytometry for Single-Cell Cellomics: High-throughput Probing Single-Cell-Resolution Signaling.

Coloring Single Nanoparticle Trajectory on Live Cell Membrane with its Own History: a Presuppositionless Approach

SCOUT: A New Algorithm for the Inference of Pseudo-Time Trajectory Using Single-Cell Data

Characterizing emerging features in cell dynamics using topological data analysis methods.

Cell-TIMP: Cellular Trajectory Inference based on Morphological Parameter

Reconstructing developmental trajectories using latent dynamical systems and time-resolved transcriptomics

Computational methods and fast hybrid stochastic simulations for triangulation sensing and identifying principles of cell navigation in the brain

Topological Data Analysis in Time Series: Temporal Filtration and Application to Single-Cell Genomics

Gene Trajectory Inference for Single-cell Data by Optimal Transport Metrics